December 4, 2025
Wrapping Up
Overview
- Conclusion
- Examples
- Exam Prep
Conclusions
Learning Objectives
- Give reasons for why truth is important in politics.
- Identify the types of claims made (implicitly or explicitly) in real-world political discourse.
- Identify which of these claims can be subject to scientific scrutiny.
- Explain what distinguishes scientific from unscientific forms of evidence.
- Analyze real-world descriptive and causal claims about politics and identify specific errors with the use of evidence and explain why these errors occurred.
- Analyze real-world causal claims and identify the types of causal evidence provided and explain the assumptions needed for this evidence to be valid.
- Apply these tools of analysis to explain how both claims and a critical engagement with the evidence for those claims change their opinion on the issue at hand.
Examples
- Should we regulate/limit the use of AI in the workplace?
- Should we regulate social media content and messages?
Why Truth?
- Instrumental rationality / clarity
- For power to not be domination: need informed consent
- Building and sharing a world together impossible if we all chase lies
How do we learn the truth?
- What kinds of claims are people making?
- Which of these claims are empirical vs. normative
How do we learn the truth?
Science:
- examines empirical claims (that speak to normative claims)
- employs evidence that is capable of showing claims to be wrong (severity)
- only works with “common sense”… sensus communis: perceptions and understandings of the world must be communicated with others
Descriptive Claims:
- Concept \(\to\) is definition systematic/transparent?
- Variable/validity \(\to\) Does variable capture concept? Does choice of variable lead us to support claim even if it is false?
- Measure/measurement error \(\to\) Do measurement strategies lead to error? What kind of error? Does this error lead us to support claim even if it is false?
Causal Claims:
- \(\color{red}{\text{Counterfactuals}}\) \(\to\) Correlation \(\to\) Confounding? \(\to\) Direction? \(\to\) Severity?
- What is the “solution” to confounding used?
- How does it “solve” confounding? (what sources of confounding are removed?)
- What assumptions does it make? (Do these assumptions seem plausible?)
- What trade offs does it make?
Practically
- Is this media or message that I am exposed to attempting to persuade me or change my behavior?
- What is this message attempting to get me to do/believe?
- How does this message attempt to persuade or motivate me?
- Does the message employ claims and evidence?
- What could go wrong with this evidence?
- What would be better evidence?
Applications
Automation and Wages
Acemoglu and Restrepo (2022) investigate:
Has automation of work helped or hurt workers?
Automation and Wages
Data:
- Exposure to automation (\(X\)): industry-level investment in robotics and software and adoption of these technologies outside the US
- Wages (\(Y\)): Hourly Real Wages for workers.
“Cases”:
- Demographic groups defined by race, age, gender, education levels
- For each group, calculate “exposure to automation” based on the industries they work in
- For each group, calculate hourly real wages
Automation and Wages
Rather than looking at wages in industries with more automation, or change in wages in the US over time, use a difference in differences:
They compare:
Change in real wages for demographic groups with high exposure to automation between 1980 and 2016 (change in \(Y\) for group where \(X\) changes)
Change in real wages for demographic groups with low/no exposure to automation between 1980 and 2016 (change in \(Y\) for group where \(X\) does not change)
Assume that counterfactual trend in wages for workers exposed to automation SAME as factual trend in wages for workers not exposed to automation.
For groups with greater increase in automation exposure, greater decline in wages
Automation and Wages
Correlation suggestions Automation \(\xrightarrow{causes}\) declining wages
- Can’t be confounding due to unchanging differences b/t demographic groups, industries
- Can’t be confounding due to factors similarly affecting all groups (e.g. national/global changes)
For this to be the causal effect of automation, need to believe that wages for workers exposed / not exposed to automation would have been similar without automation…
No differences in wage trends before automation.
Automation and Wages
It still could be that other things that affect wages changed differently for workers exposed to automation than for those who were not.
- Read the paper to see how authors rule out alternatives (e.g. moving manufacturing jobs elsewhere due to trade competition)
“capital takes what it will in the absence of constraints and technology is a tool that can be used for good or for ill… Yes, [during the Industrial Revolution of the 19th Century] you got progress, but you also had costs that were huge and very long-lasting. A hundred years of much harsher conditions for working people, lower real wages, much worse health and living conditions, less autonomy, greater hierarchy. And the reason that we came out of it wasn’t some law of economics, but rather a grass roots social struggle in which unions, more progressive politics and, ultimately, better institutions played a key role — and a redirection of technological change away from pure automation also contributed importantly.”
- Daron Acemoglu
Luddites?
Yes… Luddites.
Yes… Luddites.
